Models

Twice or four times the size, sound carried across.

The first Bria cut-out.

Up to 8K, with a preset for the kind of footage it is.

Half the wait, a little less nuance.

Quick upscaling, sound kept.

Five to twenty seconds with sound.

The same, with more care taken.

Edits a picture from four references.

The adjustable FLUX, one to four megapixels.

The slow FLUX, for the last of the detail.

FLUX at its usual best.

Quick pictures at 1K, and edits of your own.

The careful Gemini, as far as 4K.

Fourteen shapes, from half a K to four.

The cheap Gemini: 1K, four at a time.

Reads a picture and edits it as asked.

The middle GPT picture model.

Fifteen shapes, up to 4K, one at a time.

Fifteen shapes, up to 4K, four at a time.

Ten at a time, one reference in.

Six to fifteen seconds at 720p.

Seven shapes, a dozen at a time.

Five or ten seconds between two frames. No sound.

Six or ten seconds from a first frame. No sound.

The quick Hailuo 2.3.

Three to fifteen seconds from a first frame.

The newer HappyHorse.

A presenter from one photograph, up to 1080p.

Five or ten silent seconds at 720p.

1080p, an ending frame, and sound if you want it.

Three to fifteen seconds, sound optional.

Kling 3 with several pictures to work from.

The quick Kling, up to 1080p.

The first Kling avatar.

The first Kling avatar, at its better cut.

The careful Kling avatar: a picture, a recording, a talking head.

The quicker Kling avatar.

Kling's reasoning cut.

Creates a soundtrack with separate direction for effects and background music.

Cuts the objects you name — "presenter, monitor" — out of a video, each as its own clip with a transparent background, ready to layer and rearrange. Whatever crosses in front of an object is left out of its cutout. The clips carry no sound. Runs on this device.

Sharper, bigger video from the clip you have — ×4, capped at what H.264 can hold so the file plays anywhere — with the original sound carried across untouched. Runs on this device.

Follows a flat surface you mark on one frame — a screen, a poster, a frame — through the whole clip, and files where it is on every frame so anything can be laid on it. Makes no video; runs on this device.

Keeps whatever the shot is of — a person, a pet, a product — and makes everything behind it transparent, sound and all. Runs on this device.

Keeps the subject of a picture and turns everything behind it transparent. Runs on this device and returns a full-size PNG.

Writes down what is said in a clip, word by word with a time on each, and fills the clip with the captions. Runs on this device.

Four to fifteen seconds at 2K, with an audio track.

Films an avatar of your own saying whatever you hand it.

Returns several synchronized sound-effect tracks for the supplied video clips.

Creates one or more synchronized sound-effect variations for a video.

Returns multiple synchronized audio tracks from video.

Adds prompt-directed, synchronized sound effects and ambience to a video.

The lifelike one, in thirty languages.

A song from a description, up to ten minutes.

Four to ten seconds, as far as 4K.

Films the person in a picture saying what you write, at 720p or 1080p.

Reads a line out in a voice of theirs, or syncs a recording.

One to fifteen seconds; five or eight between two frames.

Seven shapes, six at a time, and a negative prompt.

The slower Qwen 2.0.

Up to three references, six pictures out.

The slower Qwen 3.0.

What is said in a clip, word by word, speaker by speaker.

Two to twelve silent seconds, quickly.

The slower Seedance 1.0.

Seedance with sound and a picture to start from.

Four to fifteen seconds with sound, as far as 4K.

The small Seedance 2.0.

Up to half a minute, with sound.

Sharp illustrations and photographs, up to 4K.

The newer Seedream, from 2K up.

Sets of up to fifteen pictures, as far as 4K.

One picture at a time, and the best of them.

Upscales to a size you name, or by a factor.

The careful SkyReels.

Three to fifteen seconds, quickly.

Returns a synchronized sound-effects track for the full joined video.

Adds synchronized sound effects to an existing video.

Up to twenty seconds with sound, 720p.

Sora at up to 1080p.

Short noises, looping if you need them to.

The newest Suno.

The smaller, quicker Suno.

Suno with a wilder creative range.

The older Sync, cheaper and quicker.

Sync's second generation.

The careful Sync 2.

The same, acted out with a feeling you pick.

Puts a recording in the mouth of whoever is on screen.

Eight seconds with sound, from a prompt or two frames.

Veo through Google's own channel: sound is a switch here.

The cheap Veo.

The slower Veo: same eight seconds, more detail.

The slower official Veo.

Three to sixteen seconds, up to 1080p.

From one second up, with smarter effects.

Vidu with dialogue and sound effects.

The quick Vidu, sound included.

Five seconds with sound.

Five, ten or fifteen seconds, sound optional.

Quick video from a picture, two to fifteen seconds.

Two to fifteen seconds, up to 1080p.

Four at a time, edits included.

The slower Wan picture model.

Builds the shot from several reference pictures.

The newest Wan, with sound.

Seven shapes in a hurry.

Increase delivery resolution while preserving the cut.

Create a separated foreground layer for compositing.

Extract a clean subject from a still image.

Transcribe speech and place captions on the timeline.

Transform an existing clip while retaining edit context.

Generate synchronized atmosphere and effects from a scene.

Reduce noise and improve spoken-word clarity.